Framing the web: cognitive modularity and the limits of belief revision
Bibliographic record
Abstract
Belief revision practices ought to respect the principles of coherence, according to standard norms of rationality.Yet numerous empirical studies suggest our belief revision practices fall hopelessly short of this goal.Worse, a number of influential accounts in cognitive science note that there are hard computational limits involved in any sort of holistic, global belief revision.We are faced with what cognitive scientists call the frame problem, which alludes to the difficult question of where to stop considering evidence before committing to (or rejecting) any given belief, yet at the same time, trapped in what Cherniak (1986) refers to as the finitary predicament of having limited time and computational resources to engage in the process.I argue that an effective way to escape this dilemma is to invoke a modular cognitive architecture, where belief revision practices are sub-served, mediated, and heavily circumscribed by informationally encapsulated cognitive mechanisms and heuristic processing.A number of influential accounts have emerged in recent years arguing for such "massively modular" systems as a response to various aspects of the frame problem (Carruthers, 2006a; Jackendoff, 2007;Sperber, 2005;Barrett & Kurzban, 2006).I defend my own version of such an account, with a specific emphasis on the question of belief revision within such a modular framework.I begin by exploring the Fodor's (1983) thesis of perceptual modularity and then elaborate the idea, arguing that there is evidence for assembled modular structures, including integrative modular assemblies that can execute belief revision processes in a computationally tractable fashion, despite Fodor's well-known objections to this extension of his theory.I describe a modular, heuristically driven cognitive system that is plausibly capable of approximating the sort of global, holistic belief revision practices that rationality demands, while maintaining computational tractability.The price of such a system, however, is that it is error-prone-it will have systematic patterns of breakdown, where some beliefs will turn to out to be essentially unrevisable and some inconsistencies of belief will be irremediable.I argue that this prediction of the account is confirmed by current research on memory distortion and delusion.Finally, I demonstrate how my account may illuminate and help resolve some ongoing debates regarding the etiology, doxastic status, and potential treatment of certain monothematic delusions.iii RésuméLes pratiques de révision des croyances doivent respecter les principes de cohérence, selon les normes de la rationalité.Pourtant, de nombreuses études empiriques suggèrent que nos pratiques de révision des croyances ne respectent clairement pas cet objectif.Pire encore, un certain nombre de théories influentes dans les sciences cognitives notent qu'il y a des limites de calcul formidables impliqués dans toute sorte de révision holistique de croyance.Nous sommes confrontés au problème de cadre, qui renvoie à la question difficile de savoir où arrêter l'examen des preuves avant de s'engager à (ou de rejeter) une croyance, mais en même temps, pris au piège dans ce que Cherniak (1986) appelle la situation finie d'ayant peu de temps et de ressources de calcul à dédier à ce processus.Je soutiens qu'un moyen efficace d'échapper à ce dilemme est d'invoquer une architecture cognitive modulaire, où les pratiques de révision des croyances sont sousdesservies et fortement encadrées par des mécanismes cognitifs encapsulés ainsi que le traitement heuristique.De nombreuses théories influentes qui ont émergé au cours des dernières années défendent les systèmes "massivement modulaires" comme réponse à divers aspects du problème de cadre (Carruthers, 2006a; Jackendoff, 2007;Sperber, 2005;Barrett & Kurzban, 2006).Je défends ma propre version d'une telle théorie, avec un accent particulier sur la question de la révision des croyances dans un cadre modulaire.Je commence par explorer la thèse de la modularité de Jerry Fodor (1983), puis j'élabore l'idée, soutenant qu'il existe des preuves de structures modulaires assemblées, y compris les assemblages modulaires intégrées qui peuvent exécuter des processus de révision des croyances dans un mode de calcul tractable, malgré les objections de Fodor à cette extension de sa théorie.Je décris un système cognitif modulaire, entraîné par des processus heuristiques, qui est probablement capable de rapprocher les pratiques de révision de croyance holistique exigées par la rationalité, tout en conservant tractabilité informatique.Toutefois, le prix d'un système de croyance est qu'il est une source d'erreurs et qu'il y aura des mouvements systématiques de rupture, où certaines croyances sont essentiellement non révisables et certaines incohérences seront irrémédiables.Je soutiens que cette prédiction de la thèse est confirmée par de nombreuses études empiriques sur les distorsions de la mémoire et croyances délirantes.Finalement, je démontre comment ma thèse peut éclairer et aider à résoudre certains débats en cours sur l'étiologie, l'état doxastique et le traitement potentiel de certains délires monothématiques.Many thanks are owed to many people who helped me to see this project through to fruition.Foremost thanks go to my supervisor, Ian Gold, whose positivity, patience, and insightful critical commentary made this thesis possible.Much gratitude is similarly owed to Jim McGilvray, whose incisive comments on earlier drafts shaped many of the ideas in this dissertation for the better.I have also had the benefit of stimulating conversations and courses with many members of the Philosophy Department, as well as my fellow graduate students.A large debt is also owed to the Department support staff, Mylissa, Angela, Claudine, and Saleema, who have been immeasurably helpful navigating through the program and the University.Finally,
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.011 | 0.043 |
| Meta-epidemiology (narrow) | 0.000 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.002 |
| Science and technology studies | 0.002 | 0.030 |
| Scholarly communication | 0.010 | 0.024 |
| Open science | 0.002 | 0.007 |
| Research integrity | 0.004 | 0.005 |
| Insufficient payload (model declined to judge) | 0.004 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".